re:Invent 2026: Bryan's schedule

Monday, November 30 to Friday, December 4. Home base: Venetian.

All times are Las Vegas time (Pacific). Reserved means a seat held in my name; Backup means a session I want without a reserved seat; Walk-in sessions need no seat.

Meet Bryan

Calendar feeds

refreshed 2026-11-06

First time at re:Invent: how to get around and what to expect

Each card answers one question and names the year its facts describe. For 2026 pickup points and shuttle times, confirm in the re:Invent app.

How do I get between venues?

Walk inside the Venetian, Caesars, and Wynn cluster. Take the monorail or a shuttle to reach MGM Grand, and use the free trams on the south Strip.

  • Monorail: A paid elevated train on the east side of the Strip from MGM Grand to SAHARA. The full 3.9-mile ride takes about 15 minutes, with trains every 4 to 8 minutes (2026 route map).
  • Free Monorail Ride: In 2024, badge holders rode free between MGM Grand Station and Harrah's/The LINQ Station, and staff checked badges at the gate. Confirm 2026 in the re:Invent app.
  • Monorail Fares: In 2026, one ride cost $6.25 at a station or $5.75 as an online eTicket. A one-day pass cost $15.50, or $13.95 online.
  • Mandalay Bay Tram: Free, with no ticket needed. It runs Mandalay Bay to Luxor to Excalibur every 3 to 7 minutes. The return from Excalibur is express to Mandalay Bay (2024 and 2026 guides).
  • Aria Express Tram: Free. It runs Park MGM to the Crystals mall (the Aria and Vdara stop) to Bellagio South Tower every 7 to 10 minutes, daily 8am to 2am (2024).
  • Not Running: The Mirage to Treasure Island tram. The Mirage closed July 17, 2024, and the tram is out of service (2026).
  • Deuce Bus: In 2023, the RTC Deuce bus ran along Las Vegas Boulevard every 10 to 15 minutes for $4 a ride. It covers hotels the shuttles skip.
Walking times between venue pairs
Walking times between re:Invent venue pairs
From and toWalkNotes
Venetian to Caesars Palace5 to 10 minPedestrian bridge
Wynn to Encore5 minSame complex, connected
Venetian to Wynn15 to 20 minConnected walkway, indoors most of the way
Venetian to Caesars Forum15 to 20 minConnected through the Expo walkway
Caesars Forum to Wynn and Encore15 to 20 minConnected walkways
MGM Grand to Mandalay Bay20 to 25 minNot walkable in practice; use the free tram or a shuttle
  • Shuttle Pairs: MGM Grand to the Venetian, Caesars, and Wynn cluster is nearly an hour on foot and 15 to 30 minutes by car. Mandalay Bay to that cluster also needs a shuttle or the monorail.
  • North to South: From north to south the venues sit in this order: Wynn and Encore; Venetian, Caesars Forum, and Caesars Palace; MGM Grand; Mandalay Bay (2024).

Source: lvmonorail.com (2026); Classmethod (2024); Vegas Unzipped (2026); Jimmy Dahlqvist (2026); AWS Builders (2023)

Where do I catch a shuttle, and when do they run?

Free shuttles connect the re:Invent venues, and a badge gets you on. AWS publishes the boarding spots and times each year, so confirm 2026 in the re:Invent app.

  • Cost: Free with a badge (2023).
  • Traffic: Shuttles share Strip traffic. Allow extra time when a session sits at a distant venue (2023).
  • MGM Grand, 2024: Ground floor of the self-parking garage, near the MGM Underground exit. MGM lists that spot for rideshare; the shuttle spot varies by year.
  • Venetian, 2024: Near the convention center entrance. The attendee guide listed the exact doors and levels.
  • Caesars Forum, 2024: Dedicated shuttle lanes outside the venue.
  • Wynn and Encore, 2024: The hotel entrance area.
  • Mandalay Bay, 2024: The hotel entrance and shuttle area.
  • Shuttle Map: The re:Invent app and reinvent.awsevents.com show color-coded routes by venue cluster, pickup and drop-off points, and estimated travel times.
  • Frequency and Hours: Set each year by route and traffic. Confirm 2026 times in the re:Invent app.
  • Evening Example: In 2024, shuttles to re:Play left every re:Invent hotel from 7pm and used the same pickup points as the session shuttles.

Source: AWS Builders (2023); MGM Resorts (2024); Classmethod re:Play (2024)

What is re:Play on Thursday night, and how do I get there?

re:Play is the official party that closes the week: live music, food, drinks, themed lounges, and games. It needs a badge. I have no confirmed 2026 venue, so confirm the venue and shuttle pickup in the re:Invent app.

re:Play venues and performers by year
YearVenuePerformers
2023Las Vegas Festival GroundsMajor Lazer, A-Trak, Portugal. The Man, King Youngblood
2024Las Vegas Festival GroundsWeezer, ZEDD
2025Reported as Allegiant Stadium; not confirmedSt. Vincent, Calvin Harris
  • 2025 Venue: The AWS homepage named Allegiant Stadium for the December 3 party, while some sources still list the Festival Grounds.
  • Shuttles, 2024: Free shuttles left every re:Invent hotel from 7pm.
  • Monorail, 2022: Ride to SAHARA Station, then walk about 10 minutes through the SAHARA casino and out to the street, following the signs to the Festival Grounds.
  • Badge Rule: Badge-only entry. Bags larger than a small purse or backpack stay out. Each person picks up their own badge with photo ID.
  • What To Bring: A jacket. The Festival Grounds are outdoors, and December nights run cold and windy (2023 and 2024).
  • Timing: In 2023 and 2024, bars closed at 11pm and the party ended at midnight.
What was there in 2023 and 2024
  • Stages and Food: A Main Stage, South and North areas, food tents, and bars.
  • Games: RC racecars, roller skating, a giant slide, dodgeball, mini golf, and arcade games.
  • Silent Disco: A silent disco with several headphone channels.

Source: AWS (2026); iret (2023); AP-Com (2024); Classmethod monorail to re:Play (2022); AWS homepage (2025)

Which session format should I pick?

Pick by what you want to do: listen, ask questions, or build. Workshops and builders' sessions need a laptop. The other formats do not.

  • Workshop: 2 hours. You build a solution in an AWS account with an expert guiding you. Bring a laptop. Not recorded. Suits you if you learn by doing.
  • Builders' Session: 90 minutes. A small group codes or designs with an expert at your side. Bring a laptop. Not recorded. Suits you if you want help on a specific build.
  • Chalk Talk: 60 minutes. A small group works through an architecture on a whiteboard with experts. No laptop. Not recorded. Suits you if you have a design question to ask.
  • Breakout Session: 60 minutes. A lecture by AWS experts, customers, or partners, with 10 to 15 minutes of questions. No laptop. Recorded, and posted to YouTube after the event. Suits you if you want an overview and can watch the recording later.
  • Code Talk: 60 minutes. Live coding or a code walkthrough. No laptop. Recording varies by session listing.
  • Lightning Talk: 15 to 20 minutes. A short talk on one focused topic. No laptop. Recording varies.
Session levels, 100 to 500
  • 100, Foundational: Getting started, for beginners who want a solid overview.
  • 200, Intermediate: Building the foundations, for people ready to go deeper.
  • 300, Advanced: Growing expertise, with a comprehensive dive into selected topics.
  • 400, Expert: Advanced implementation, for people who want the intricacies.
  • 500, Distinguished: The highest level, for experts pushing boundaries.
Reserved seats, walk-up seats, and the 10-minute seat release
  • Reserved Seats: About 75% of seats can be reserved in advance. Reservations typically open in mid-October. Reserving is separate from marking a session as a favorite (2026).
  • Walk-Up Seats: The remaining 25% or so fill from the walk-up line on the day (2026).
  • Seat Release: In 2023, reserved seat holders had to be scanned in about 10 to 15 minutes before the start. After that the seat went to the walk-up line.
  • Queue Timing: In 2023, lines formed about 60 minutes ahead, reserved holders entered from about 30 minutes ahead, and the walk-up line entered at about 10 minutes ahead.
  • Keynotes: Walk-up only, first come first served, with no reserved seats (2026).
  • Repeats: Popular sessions repeat at other times and venues. Check the catalog for a repeat if you miss one.

Source: AWS, how you will learn (2026); AWS Heroes (2024); AWS Builder Center (2023); Jimmy Dahlqvist (2026); AWS Builders queueing guide (2023)

Where do I pick up my badge, and what do I do the first morning?

Pick up your badge before Monday if you can. Bring photo ID, because each person collects their own badge.

  • Where: In 2024 and 2025: Harry Reid Airport, the Venetian, MGM Grand, and Mandalay Bay. Confirm 2026 locations in the re:Invent app.
  • Photo ID: Required at pickup. There is no bulk pickup (2024).
  • Sunday Pickup: Collecting a badge on Sunday avoids the Monday morning lines (2024).
  • Swag: In 2024, the swag desk sat at the Venetian behind badge pickup. Try on hoodies before you queue, because popular sizes run out.
  • The App: Download the re:Invent app. Your reserved sessions, room locations, and shuttle maps live there (2026).
  • Monday Evening: The Welcome Reception, around 6pm, gives early access to the Expo (2026).
  • Content Hubs: Each venue has one, with simulcasts, self-paced labs, and meetups. Use one as a backup when a session is full (2026).
  • Meals: In 2023, breakfast ran 6am to 8:30 or 9am and lunch ran 11am to 1:30pm, both with strict end times. Dinner was not included.
2024 badge pickup hours by location
2024 badge pickup hours by location
LocationSundayMondayTuesday to Thursday
Harry Reid Airport, Terminals 1 and 37am to midnight7am to midnightClosed
Venetian, Level 2, Venetian Ballroom F10am to 8pm6:30am to 9pm6:30am to 8pm
MGM Grand10am to 8pm6:30am to 9pm6:30am to 8pm
Mandalay Bay10am to 8pm6:30am to 9pm6:30am to 8pm

Source: Bespin Global (2024); Classmethod badge pickup (2024); Labra (2025); AWS, know before you go (2026); AWS Builders (2023)

Hallway rules
  • Leave session feedback as soon as the session ends. Every session has feedback available right after; do it before the next room, while it is fresh.
  • Thank the host in the hallway.
  • Be polite to the staff. They are Las Vegas citizens turning the room for the next session, one of dozens of shows a year.
  • Get water and keep moving. There is a hydrate-and-help challenge running all week; the hydration stations are on the campus map in the event app.

Where do I eat, and what happens in the evenings?

Breakfast and lunch are served on-prem at the session venues: eat where your sessions are. Breakfast is grab-and-go or sit-down, depending on whether the gap before the session is longer than 20 minutes. Grab-and-go boxes cover sessions over the window; confirm 2026 times in the event app on site. Dinners are not served on-prem.

Evenings

The official evening lineup, quoted from the event site. re:Play on Thursday is my one after-hours event; the other nights I am out.

  • Monday: Expo Welcome reception: drinks, early Expo floor access, first conversations of the week.

  • Tuesday: Trivia Night with Jeff Barr, Game Night, Dungeons and Dragons miniature mixer, plus partner networking cocktail hours.

  • Wednesday: re:Invent Run at Allegiant Stadium (supports the Michael J. Fox Foundation), Expo Happy Hour, Magic: The Gathering night, plus partner networking cocktail hours.

  • Thursday: re:Play, the official re:Invent party: live performances, food, games, themed lounges. Complimentary shuttles from re:Invent hotels.

Monorail stops, south to north
  • MGM Grand
  • Bally's / Paris Las Vegas
  • Flamingo / Caesars Palace
  • Harrah's / The LINQ
  • Las Vegas Convention Center
  • Westgate
  • Sahara

Station markers sit at the named property and are approximate.

What re:Invent looked like in 2025

Three photos from the re:Invent 2025 floor, December 1
Caesars Forum, convention center, re:Invent 2025
Three photos from the re:Invent 2025 floor, December 1
Caesars Forum and Venetian Expo, re:Invent 2025
Three photos from the Sports Forum at re:Invent 2025
Sports Forum, re:Invent 2025
Three photos from the re:Invent 2025 floor, December 2
Caesars Forum, convention center, re:Invent 2025
Three photos from the re:Invent 2025 floor, December 3
MGM Grand Conference Center, re:Invent 2025
Three photos from the re:Invent 2025 floor, December 3 and 4
Venetian, re:Invent 2025
Three photos from the re:Invent 2025 floor, December 4
Venetian, resort, re:Invent 2025
Three photos from the re:Invent 2025 floor, December 4 and 5
Caesars Forum and Venetian Expo, re:Invent 2025
Three photos from re:Play at re:Invent 2025
re:Play, Thursday night, re:Invent 2025
Three photos from the re:Invent 2025 floor, December 5
re:Play and Venetian Expo, re:Invent 2025
Three photos from the re:Invent 2025 floor, December 5 and 6
re:Play, Venetian Expo and Palazzo Tower, re:Invent 2025

Daily Schedule

Where I will be each day, in order, with times, rooms, and status. Open windows and the Meet Bryan list come after Friday.

Monday, November 30

1 venue · about 30 min in transit

Main venue: MGM Grand, 4 reserved sessions

Monday, November 30 at a glance
TimeWhatWhereStatus
morning, before 08:30BreakfastMGM GrandMeal
08:30–09:30STG363 Modernizing VMware workloads with Amazon FSx for NetApp ONTAPMGM Grand, Level 3, Room 306Reserved
08:30–09:30AIM401-R Agents that train themselves: autonomous reinforcement learning on AWSMGM Grand, Level 3, Chairman's 363, Content Hub, Builders' Session 2Backup
09:30Walk to Level 3, Room 350MGM GrandTravel
09:30–11:30Open window, lunchMGM GrandOpen
10:00–11:00AIM346-R Deploy, tune, and scale AI workloads on AWS with self-managed agentsMGM Grand, Level 1, Boulevard 167Backup
11:30–12:30INV530-R1 EC2 at 20: Rebuilding the control plane while it is runningMGM Grand, Level 3, Room 350Reserved
12:00–13:00CMP328-R Deploy AI agent sandboxes on EC2MGM Grand, Level 1, Boulevard 155Backup
12:30Walk to Level 1, Boulevard 163MGM GrandTravel
12:30–13:00LunchMGM GrandMeal
13:00–14:00CMP340 Drug discovery and materials use cases with Amazon Braket and Q-CTRLMGM Grand, Level 1, Boulevard 163Reserved
13:00–14:00AIM312-R Beyond fine-tuning: RL for reasoning agents with AWS TrainiumMGM Grand, Level 3, Premier 320Backup
14:00Walk to Level 1, Terrace 153MGM GrandTravel
14:00–15:00Open midday windowOpen
15:00–17:00AIM318 Resilient VLM Training at Scale on Amazon SageMaker HyperPodMGM Grand, Level 1, Terrace 153Reserved
15:00–16:00NET337 Extending your cloud network with Amazon Leo satellite connectivityMGM Grand, Level 3, Room 353Backup
16:00–17:00NET335 Building searchable live sports with AWS Elemental and Amazon CloudFrontMGM Grand, Level 1, Grand 117Walk-in
18:00–18:20COM344 Now serverless has servers. What now?Venetian, Level 2, Hall B, Expo, Builder TheaterWalk-in

Details

morning, before 08:30
breakfast · MGM Grandbreakfast near MGM Grand — grab-and-go or sit-down by the 20-minute gap; room and serving hours posted on site
First leg: shuttle from Venetian to MGM Grand, Level 3, Room 306. Map of the rooms for the day
Schematic map of MGM Grand, 4 numbered rooms in visit order: 1 Level 3, Room 306; 2 Level 3, Room 350; 3 Level 1, Boulevard 163; 4 Level 1, Terrace 153
  1. 08:30-09:30 STG363 Modernizing VMware workloads with Amazon FSx for NetApp ONTAP Level 3 | Room 306
  2. 11:30-12:30 INV530-R1 EC2 at 20: Rebuilding the control plane while it is running Level 3 | Room 350
  3. 13:00-14:00 CMP340 Drug discovery and materials use cases with Amazon Braket and Q-CTRL Level 1 | Boulevard 163
  4. 15:00-17:00 AIM318 Resilient VLM Training at Scale on Amazon SageMaker HyperPod Level 1 | Terrace 153

Indoor layout is a schematic, not to scale.

(c) OpenStreetMap contributors

08:30–09:30 STG363Chalk talkLevel 300 – Advanced

Modernizing VMware workloads with Amazon FSx for NetApp ONTAP

John JamailSatya Vandrangi

Migrating VMware workloads to AWS often means rethinking storage, but it doesn't have to. In this chalk talk, we explore how Amazon FSx for NetApp ONTAP provides a storage foundation built for VMware migrations.

Read the rest

Learn how FSx for ONTAP delivers multi-protocol support, enterprise data management, and ONTAP compatibility that VMware administrators already know. We'll discuss real world migration patterns, cost optimization through deduplication and compression, and high availability with Multi-AZ and zero RPO. Bring your questions and join us to design a migration strategy that minimizes disruption and maximizes your existing ONTAP investments.

MGM Grand, Level 3, Room 306

09:30–11:30
open window · lunch near MGM Grandmorning window120-minute window near MGM Grand — grab-and-go or sit-down by the 20-minute rule

Next: walk to MGM Grand, Level 3, Room 350, about 10 min

11:30–12:30 INV530-R1Chalk talkLevel 500 – Distinguished

EC2 at 20: Rebuilding the control plane while it is running

Don McGarry

Amazon EC2 launched in 2006 as four services backed by a single MySQL database. Twenty years later, it is one of the largest distributed systems ever built — and most of that original architecture had to be replaced while serving production traffic, with no maintenance window and no change to the API contract customers depend on.

Read the rest

In this session, EC2 engineers dive deep on two of the hardest migrations: making the control plane Availability Zone independent so a single-AZ event cannot escalate into a regional one, and moving EC2 metadata off MySQL. We walk through the failure modes that forced each decision, how correctness was verified before every cutover, how we proved a single-AZ failure stays in a single AZ, and the things we got wrong along the way. You will leave with migration and isolation patterns you can apply to your own systems at any scale.

MGM Grand, Level 3, Room 350

12:30–13:00
lunch · MGM Grand30-minute window near MGM Grand — grab-and-go or sit-down by the 20-minute rule

Next: walk to MGM Grand, Level 1, Boulevard 163, about 10 min

13:00–14:00 CMP340Chalk talkLevel 300 – Advanced

Drug discovery and materials use cases with Amazon Braket and Q-CTRL

Nick White

Classically hard problems such as simulating strongly correlated electrons and 1D NMR spectra, critical to material and drug discovery, can be sped up on quantum computers but until now had not been demonstrated. Q-CTRL shows how Fire Opal on Amazon Braket enables simulating 1D NMR spectra for spin systems up to 34 qubits and achieved a 3,000x speedup on the Fermi-Hubbard model at 120 qubits, demonstrating practical quantum advantage.

Read the rest

This chalk talk unpacks how AWS hybrid workflows enabled these outcomes and what enterprises should do now to capture ROI with quantum.

MGM Grand, Level 1, Boulevard 163

14:00–15:00
Open midday window

Next: walk to MGM Grand, Level 1, Terrace 153, about 10 min

15:00–17:00 AIM318WorkshopLevel 300 – Advanced

Resilient VLM Training at Scale on Amazon SageMaker HyperPod

Roy AllelaAravind Neelakantan

Training vision-language models at scale fails in predictable ways — node crashes mid-epoch, checkpoints corrupt, and GPU clusters sit idle during recovery. In this hands-on workshop, build resilient VLM training pipelines on Amazon SageMaker HyperPod that survive real-world failures.

Read the rest

Configure automatic health monitoring and node replacement, implement fault-tolerant checkpointing strategies, and design distributed training topologies that resume without losing progress. Using open-source frameworks on multi-node GPU clusters, you'll train a VLM end-to-end while simulating and recovering from infrastructure failures. Leave with battle-tested patterns for keeping large-scale training runs alive.

MGM Grand, Level 1, Terrace 153

Tuesday, December 1

2 venues · about 55 min in transit

Main venue: Caesars Forum, 2 reserved sessions

Tuesday, December 1 at a glance
TimeWhatWhereStatus
morning, before 12:00BreakfastWynn/EncoreMeal
10:30–10:50IND105-S Amazon Leo: Connectivity Beyond the Reach of Existing Networks (sponsored by Amazon Leo)Venetian, Level 2, Hall B, Expo, Industry TheaterWalk-in
11:00–11:20BIZ111-S Build the Agentic Workplace: What's New on the Slack Developer Platform (sponsored by Salesforce)Venetian, Level 2, Hall B, Expo, Builder TheaterWalk-in
12:00–13:00IND359-R1 In-Country HPC: How TGS Processes Seismic Data on AWS BrazilWynn/Encore, Level 1, Chopin 4Reserved
12:30–14:30HMC403-R1 Self-managed agentic AI at the hybrid edgeMGM Grand, Level 1, Grand 124Backup
13:00Shuttle to Caesars ForumWynn/Encore to Caesars ForumTravel
13:00–15:00Open window, lunchCaesars ForumOpen
14:30–15:30OPN406-R1 Cut token costs by 50%: Building context efficient agents with StrandsMGM Grand, Level 1, Boulevard 159Backup
15:00–16:00IND322-R Physical AI in Telecom: A Framework for Workload PlacementCaesars Forum, Level 1, Academy 414Reserved
16:00Walk to Level 1, Forum 114Caesars ForumTravel
16:00–17:00CMP308 Modernizing HPC on AWS: Migration at scale with PCSMGM Grand, Level 3, Chairman's 363, Content Hub, Red TheaterWalk-in
16:30–17:30IND3304-R1 KNFSD File Cache: Extending Your NFS Storage into the CloudCaesars Forum, Level 1, Forum 114Reserved

Details

morning, before 12:00
breakfast · Wynn/Encorebreakfast near Wynn/Encore — grab-and-go or sit-down by the 20-minute gap; room and serving hours posted on site
First leg: shuttle from Venetian to Wynn/Encore, Level 1, Chopin 4. Map of the rooms for the day
Schematic map of Wynn/Encore, 1 numbered room in visit order: 1 Level 1, Chopin 4
  1. 12:00-13:00 IND359-R1 In-Country HPC: How TGS Processes Seismic Data on AWS Brazil Level 1 | Chopin 4

Indoor layout is a schematic, not to scale.

(c) OpenStreetMap contributors

12:00–13:00 IND359-R1Chalk talkLevel 300 – Advanced

In-Country HPC: How TGS Processes Seismic Data on AWS Brazil

Kun JiaoFaisal Asif

TGS -- one of the world's largest multi-client geoscience data providers -- faced a significant cost challenge running its business-critical Rackmon EFWI GPU workload at scale. By strategically deploying this workload on AWS infrastructure in Brazil, TGS leveraged Brazil's ReData Special Tax Regime to dramatically reduce the tax burden on GPU-intensive seismic processing.

Read the rest

The result: material cost savings on one of the most computationally demanding workloads in the energy industry -- and a reusable in-country processing architecture that can serve future customers with data residency requirements in other geographies.

Wynn/Encore, Level 1, Chopin 4

13:00–15:00
open window · lunch near Caesars Forummidday window120-minute window near Caesars Forum — grab-and-go or sit-down by the 20-minute rule
Next: shuttle to Caesars Forum, Level 1, Academy 414, about 45 min (estimate)
Map from Wynn/Encore (1) to Caesars Forum (2) showing both building footprints
  1. Leave Wynn/Encore
  2. Arrive at Caesars Forum (nearest monorail station: Harrah's / The LINQ)

(c) OpenStreetMap contributors

Schematic map of Caesars Forum, 2 numbered rooms in visit order: 1 Level 1, Academy 414; 2 Level 1, Forum 114
  1. 15:00-16:00 IND322-R Physical AI in Telecom: A Framework for Workload Placement Level 1 | Academy 414
  2. 16:30-17:30 IND3304-R1 KNFSD File Cache: Extending Your NFS Storage into the Cloud Level 1 | Forum 114

Indoor layout is a schematic, not to scale.

(c) OpenStreetMap contributors

15:00–16:00 IND322-RChalk talkLevel 300 – Advanced

Physical AI in Telecom: A Framework for Workload Placement

Young JungAngelo SampietroUmer Chaudhary

Some physical AI workloads need sub-millisecond performance at edge cell sites, others are better suited to a full AWS Region, and many fall in between on AWS Local Zones or AWS Outposts. In this chalk talk, we provide a framework for matching each workload to the right option.

Read the rest

We trace the physical AI development lifecycle, examining how data sovereignty, latency, data gravity, and token economics interact to shape placement decisions at each phase. Based on real-world use cases, we discuss the options so you leave with a placement framework you can apply immediately.

Caesars Forum, Level 1, Academy 414

Next: walk to Caesars Forum, Level 1, Forum 114, about 10 min

16:30–17:30 IND3304-R1Chalk talkLevel 300 – Advanced

KNFSD File Cache: Extending Your NFS Storage into the Cloud

Andy HayesDJ Rahming

Need to cache and serve NFS data from on-premises filers, across Availability Zones and Regions, or from other cloud providers with full read and write-back support? In this chalk talk, dive into KNFSD File Cache, an open-source, kernel-based NFS caching proxy on Amazon EC2 that transparently extends existing NFS storage into AWS.

Read the rest

A single Terraform module stands up a production-ready, autoscaling cluster with high throughput and at low cost. Discuss hybrid and burst-compute patterns across M&E, HPC, Financial Services, and Life Sciences industries—eliminating latency without migration, client changes, or vendor lock-in.

Caesars Forum, Level 1, Forum 114

Wednesday, December 2

2 venues · about 2 h in transit

Main venue: MGM Grand, 2 reserved sessions

Wednesday, December 2 at a glance
TimeWhatWhereStatus
morning, before 08:00BreakfastMGM GrandMeal
08:00–10:00ANT402-R Reduce spend and data movement with Amazon Redshift Data Sharing and GravitonMGM Grand, Level 1, Grand 113Reserved
08:30–09:30IND3351 AT&T Expands Connectivity with AWS Interconnect and Amazon LeoCaesars Forum, Level 1, Forum 120, Content Hub, Orange TheaterWalk-in
10:00Shuttle to Caesars ForumMGM Grand to Caesars ForumTravel
10:00–13:30Open window, lunchCaesars ForumOpen
10:00–11:00IND304 How Netflix auto-selects the right storage for every ML training workloadCaesars Forum, Level 1, Forum 120, Content Hub, Red TheaterWalk-in
12:00–13:00CMP349 Your next Quantum developer might be an AI agentMGM Grand, Level 1, Boulevard 165Backup
13:30–14:30INV537 No torque sensors, no problem: learning safe physical interaction for humanoidsCaesars Forum, Level 1, Alliance 305Reserved
13:30–14:30AIM407-R1 Build a production multi-Agent Systems with StrandsMGM Grand, Level 1, Room 104Backup
14:30Shuttle to MGM GrandCaesars Forum to MGM GrandTravel
14:30–15:30Open midday windowOpen
15:30–17:30AIM434-R1 From raw data to custom model: Dataset prep with Amazon SageMaker AIMGM Grand, Level 1, Grand 111Reserved
15:30–15:50NET341 Building hybrid satellite and cloud networks with Amazon LeoVenetian, Level 2, Hall B, Expo, Industry TheaterWalk-in
17:30–17:50AIM393 Open weights to production: How Salesforce built and served Koa modelsVenetian, Level 2, Hall B, Expo, Builder TheaterWalk-in

Details

morning, before 08:00
breakfast · MGM Grandbreakfast near MGM Grand — grab-and-go or sit-down by the 20-minute gap; room and serving hours posted on site
Start: MGM Grand, Level 1, Grand 113. Map of the rooms for the day
Schematic map of MGM Grand, 1 numbered room in visit order: 1 Level 1, Grand 113
  1. 08:00-10:00 ANT402-R Reduce spend and data movement with Amazon Redshift Data Sharing and Graviton Level 1 | Grand 113

Indoor layout is a schematic, not to scale.

(c) OpenStreetMap contributors

08:00–10:00 ANT402-RWorkshopLevel 400 – Expert

Reduce spend and data movement with Amazon Redshift Data Sharing and Graviton

Erika NagaminePoulomi Dasgupta

Organizations sharing data across business units often rely on slow, error-prone ETL jobs that create costly duplication. Amazon Redshift Data Sharing enables secure, live access across provisioned and serverless warehouses — within or across accounts and regions — eliminating data movement entirely.

Read the rest

Additionally, Graviton-powered Redshift RG instances accelerate workloads on Apache Iceberg tables through optimized columnar processing and parallel I/O. In this session, learn how to break down data silos with cross-warehouse data sharing and boost open data lakehouse performance with Redshift on Graviton.

MGM Grand, Level 1, Grand 113

10:00–13:30
open window · lunch near Caesars Forummorning window210-minute window near Caesars Forum — grab-and-go or sit-down by the 20-minute rule
Next: shuttle to Caesars Forum, Level 1, Alliance 305, about 60 min (estimate)
Map from MGM Grand (1) to Caesars Forum (2) showing both building footprints
  1. Leave MGM Grand (nearest monorail station: MGM Grand)
  2. Arrive at Caesars Forum (nearest monorail station: Harrah's / The LINQ)

(c) OpenStreetMap contributors

Schematic map of Caesars Forum, 1 numbered room in visit order: 1 Level 1, Alliance 305
  1. 13:30-14:30 INV537 No torque sensors, no problem: learning safe physical interaction for humanoids Level 1 | Alliance 305

Indoor layout is a schematic, not to scale.

(c) OpenStreetMap contributors

13:30–14:30 INV537Chalk talkLevel 500 – Distinguished

No torque sensors, no problem: learning safe physical interaction for humanoids

Josh Merel

How do you make a humanoid robot safe to operate alongside people, without the torque sensors? Fauna Robotics Sprout is a humanoid robot that uses a position-command interface, yet it must regulate the forces it exchanges with the world.

Read the rest

We show how Sprout learns compliant behavior that approximates classic impedance control, through mechanisms that control force at the joints and in task space. We describe how these controllers are trained and how they enable safer, whole-body interaction.

Caesars Forum, Level 1, Alliance 305

14:30–15:30
Open midday window
Next: shuttle to MGM Grand, Level 1, Grand 111, about 60 min (estimate)
Map from Caesars Forum (1) to MGM Grand (2) showing both building footprints
  1. Leave Caesars Forum (nearest monorail station: Harrah's / The LINQ)
  2. Arrive at MGM Grand (nearest monorail station: MGM Grand)

(c) OpenStreetMap contributors

Schematic map of MGM Grand, 1 numbered room in visit order: 1 Level 1, Grand 111
  1. 15:30-17:30 AIM434-R1 From raw data to custom model: Dataset prep with Amazon SageMaker AI Level 1 | Grand 111

Indoor layout is a schematic, not to scale.

(c) OpenStreetMap contributors

15:30–17:30 AIM434-R1WorkshopLevel 400 – Expert

From raw data to custom model: Dataset prep with Amazon SageMaker AI

gzappia ZappiaAmin Dashti

Great models start with great data. In this hands-on workshop, use SageMaker AI Agent skills to profile, curate, and validate datasets for the full model customization lifecycle — no infrastructure to manage.

Read the rest

Assess data completeness and distribution, apply auto-labeling at scale, run human-in-the-loop review for edge cases, and fix quality issues like deduplication and toxicity filtering. Generate synthetic data to fill coverage gaps, then prepare datasets for Supervised Fine-Tuning, Direct Preference Optimization, Reinforcement Learning with Verifiable Rewards, and RLAIF. Launch training, evaluate with LLM-as-a-Judge, and deploy. Walk away ready to go from raw data to custom model in hours.

MGM Grand, Level 1, Grand 111

Thursday, December 3

3 venues · about 1 h 45 min in transit

Main venue: MGM Grand, 1 reserved session

Thursday, December 3 at a glance
TimeWhatWhereStatus
morning, before 08:00BreakfastMGM GrandMeal
08:00–10:00STG412 Find and fix storage bottlenecks for self-managed OLTP on EC2 with EBSMGM Grand, Level 1, Grand 120Reserved
08:00–10:00AIM402-R From dataset to a deployed fine-tuned LLM in hours on SageMaker AIMGM Grand, Level 3, Premier 317Backup
10:00Shuttle to Wynn/EncoreMGM Grand to Wynn/EncoreTravel
10:00–12:00Open window, lunchWynn/EncoreOpen
10:30–11:30ARC322 Resilience patterns for the disconnected edge with Amazon LeoMGM Grand, Level 1, Grand 115Walk-in
11:30–12:30AIM405 Build agents on OSS models with blazing fast inference on SageMaker AIMGM Grand, Level 1, Boulevard 158Backup
12:00–14:00SEC401 Winning the DDoS Game: defend your app under real-time attackWynn/Encore, Upper Convention Promenade, Cristal 5Reserved
13:00–14:00CMP321-R1 Build an agentic AI system on Amazon EKS with CPU-only inferenceMGM Grand, Level 1, Boulevard 167Backup
14:00Shuttle to Caesars PalaceWynn/Encore to Caesars PalaceTravel
14:00–15:30Open midday windowOpen
15:00–16:00OPN312 Hugging Face on AWS: AI Model Deployment at ScaleCaesars Forum, Level 1, Alliance 305Backup
15:30–17:30MAM416 Free your .NET app from Windows licensing with AI-powered portabilityCaesars Palace, Promenade Level, Milano IVReserved
16:00–17:00SVS332-R Lambda Managed Instances: Always-on Serverless ComputeCaesars Forum, Level 1, Academy 417Backup
eveningre:PlayBus from Venetian Expo Hall CEvening

Details

morning, before 08:00
breakfast · MGM Grandbreakfast near MGM Grand — grab-and-go or sit-down by the 20-minute gap; room and serving hours posted on site
Start: MGM Grand, Level 1, Grand 120. Map of the rooms for the day
Schematic map of MGM Grand, 1 numbered room in visit order: 1 Level 1, Grand 120
  1. 08:00-10:00 STG412 Find and fix storage bottlenecks for self-managed OLTP on EC2 with EBS Level 1 | Grand 120

Indoor layout is a schematic, not to scale.

(c) OpenStreetMap contributors

08:00–10:00 STG412WorkshopLevel 400 – Expert

Find and fix storage bottlenecks for self-managed OLTP on EC2 with EBS

Anthony ThunyinShiju Purushothaman

​​When you run a self-managed OLTP database on Amazon EC2, the storage bottleneck is rarely where you assume. This hands-on session teaches a repeatable method for finding and fixing it.

Read the rest

Use Flexible IO tester and a PostgreSQL/pgbench workload to connect storage metrics to database symptoms. Measure latency, IOPS, throughput, and queue depth; determine whether the limit is the volume, instance bandwidth, or I/O pattern; then tune the right EBS volume configuration. Validate every change with EBS detailed performance statistics and CloudWatch. Leave with a proven sizing-and-tuning method for database workloads on EC2.

MGM Grand, Level 1, Grand 120

10:00–12:00
open window · lunch near Wynn/Encoremorning window120-minute window near Wynn/Encore — grab-and-go or sit-down by the 20-minute rule
Next: shuttle to Wynn/Encore, Upper Convention Promenade, Cristal 5, about 60 min (estimate)
Map from MGM Grand (1) to Wynn/Encore (2) showing both building footprints
  1. Leave MGM Grand (nearest monorail station: MGM Grand)
  2. Arrive at Wynn/Encore

(c) OpenStreetMap contributors

Schematic map of Wynn/Encore, 1 numbered room in visit order: 1 Upper Convention Promenade, Cristal 5
  1. 12:00-14:00 SEC401 Winning the DDoS Game: defend your app under real-time attack Upper Convention Promenade | Cristal 5

Indoor layout is a schematic, not to scale.

(c) OpenStreetMap contributors

12:00–14:00 SEC401WorkshopLevel 400 – Expert

Winning the DDoS Game: defend your app under real-time attack

Jonathan WoodsJulian Garcia

Compete in a team-based environment where you face real-time threats aimed at your web application. Use the latest capabilities in AWS WAF and AWS Shield to mitigate HTTP floods, web exploits, and malicious bots that disrupt your customers.

Read the rest

This risk-free event puts teamwork to the test as you work under pressure. Watch a real-time scoreboard as your team works through each level. Whether it's volumetric attacks or the new wave of AI-driven traffic hitting your APIs, you'll leave prepared to defend against what's coming next. The most effective team earns a prize!

Wynn/Encore, Upper Convention Promenade, Cristal 5

14:00–15:30
Open midday window
Next: shuttle to Caesars Palace, Promenade Level, Milano IV, about 45 min (estimate)
Map from Wynn/Encore (1) to Caesars Palace (2) showing both building footprints
  1. Leave Wynn/Encore
  2. Arrive at Caesars Palace (nearest monorail station: Flamingo / Caesars Palace)

(c) OpenStreetMap contributors

Schematic map of Caesars Palace, 1 numbered room in visit order: 1 Promenade Level, Milano IV
  1. 15:30-17:30 MAM416 Free your .NET app from Windows licensing with AI-powered portability Promenade Level | Milano IV

Indoor layout is a schematic, not to scale.

(c) OpenStreetMap contributors

15:30–17:30 MAM416WorkshopLevel 400 – Expert

Free your .NET app from Windows licensing with AI-powered portability

Mangesh BudkuleKhurram Khawaja

This workshop walks you through modernizing a .NET Framework 4.8 Web Forms application to cross-platform .NET 10. You will learn how to transform a .NET Framework 4.8 Web Forms application to .NET 10 Blazor using AWS Transform Interactive Mode.

Read the rest

Throughout the session, you will interact with the AI agent in real-time asking questions, reviewing assessment and transformation plans, and examining transformation results, including code diffs and build verification. Finally, use Kiro to identify and fix post-transformation runtime issues, such as layout problems, and containerize and deploy the transformed application to Amazon ECS Fargate using the Kiro IDE.

Caesars Palace, Promenade Level, Milano IV

Friday, December 4

1 venue · no transit time listed

Main venue: Wynn/Encore, 1 reserved session

Friday, December 4 at a glance
TimeWhatWhereStatus
morning, before 08:00BreakfastWynn/EncoreMeal
08:00–10:00BIZ321-R1 Never Start Over: Chat, Voice, Video, Screen Share on Connect CustomerWynn/Encore, Upper Convention Promenade, Cristal 1Reserved
08:00–10:00CMP318-R1 Build your iOS CI/CD pipeline with AI agents on EC2 MacWynn/Encore, Upper Convention Promenade, Cristal 5Backup
11:30–12:30CMP348-R1 Building better quantum computers with digital twinsWynn/Encore, Convention Promenade, Montrachet 1Backup

Details

morning, before 08:00
breakfast · Wynn/Encorebreakfast near Wynn/Encore — grab-and-go or sit-down by the 20-minute gap; room and serving hours posted on site
First leg: travel from Venetian to Wynn/Encore, Upper Convention Promenade, Cristal 1. Map of the rooms for the day
Schematic map of Wynn/Encore, 1 numbered room in visit order: 1 Upper Convention Promenade, Cristal 1
  1. 08:00-10:00 BIZ321-R1 Never Start Over: Chat, Voice, Video, Screen Share on Connect Customer Upper Convention Promenade | Cristal 1

Indoor layout is a schematic, not to scale.

(c) OpenStreetMap contributors

08:00–10:00 BIZ321-R1WorkshopLevel 300 – Advanced

Never Start Over: Chat, Voice, Video, Screen Share on Connect Customer

Ying QianAmit Gupta

Your customer is halfway through an application in your app, and stuck. Today they abandon it, call you, and start explaining from scratch.

Read the rest

In this hands-on workshop you build a single unbroken conversation that escalates in richness without losing context. A customer opens chat inside your application. When they stall on a form, the agent co-browses the page and sees the blocking field, then uses agent assist to answer a policy question with cited sources. The conversation carries into an in-app call, context intact, no phone number and no repeated details. Documents are reviewed over screen share, identity is confirmed over video, and because it's a joint application the co-applicant and a product specialist join the same call. Walk away with an escalation pattern combining persistent context, WebRTC channels and AI-assisted agents, for any high-consideration interaction.

Wynn/Encore, Upper Convention Promenade, Cristal 1

Which sessions still have seats?

No seat needed

Meet Bryan

My open windows this week, with where I will be.

  • Monday, November 30
    • 09:30–11:30 · MGM Grand (morning, 120 min)
    • 14:00–15:00 · MGM Grand (midday, 60 min)
  • Tuesday, December 1
    • 13:00–15:00 · Caesars Forum (midday, 120 min)
  • Wednesday, December 2
    • 10:00–13:30 · Caesars Forum (morning, 210 min)
    • 14:30–15:30 · in transit, Caesars Forum to MGM Grand (midday, 60 min)
  • Thursday, December 3
    • 10:00–12:00 · Wynn/Encore (morning, 120 min)
    • 14:00–15:30 · in transit, Wynn/Encore to Caesars Palace (midday, 90 min)

Connect on LinkedIn with "re:Invent" in the note.

Work with Bryan

  • Readiness architecture and blueprint for GenAI and RAG on AWS: Bedrock, vector search, serverless patterns, integration roadmap.

    Book the RAG blueprint
  • Two days of focused CTO-level work: incident remediation, architecture changes, or prioritized delivery for your team.

    Book CTO time
  • Full access audit of your AWS account with prioritized remediation.

    Book an access audit

Which sessions are at each venue?

MGM Grand (7)
  • STG363 · Chalk talk — Modernizing VMware workloads with Amazon FSx for NetApp ONTAP

    2026-11-30 08:30–09:30, Level 3, Room 306 · John Jamail, Satya Vandrangi
  • INV530-R1 · Chalk talk — EC2 at 20: Rebuilding the control plane while it is running

    2026-11-30 11:30–12:30, Level 3, Room 350 · Don McGarry
  • CMP340 · Chalk talk — Drug discovery and materials use cases with Amazon Braket and Q-CTRL

    2026-11-30 13:00–14:00, Level 1, Boulevard 163 · Nick White
  • AIM318 · Workshop — Resilient VLM Training at Scale on Amazon SageMaker HyperPod

    2026-11-30 15:00–17:00, Level 1, Terrace 153 · Roy Allela, Aravind Neelakantan
  • ANT402-R · Workshop — Reduce spend and data movement with Amazon Redshift Data Sharing and Graviton

    2026-12-02 08:00–10:00, Level 1, Grand 113 · Erika Nagamine, Poulomi Dasgupta
  • AIM434-R1 · Workshop — From raw data to custom model: Dataset prep with Amazon SageMaker AI

    2026-12-02 15:30–17:30, Level 1, Grand 111 · gzappia Zappia, Amin Dashti
  • STG412 · Workshop — Find and fix storage bottlenecks for self-managed OLTP on EC2 with EBS

    2026-12-03 08:00–10:00, Level 1, Grand 120 · Anthony Thunyin, Shiju Purushothaman
Caesars Palace (1)
  • MAM416 · Workshop — Free your .NET app from Windows licensing with AI-powered portability

    2026-12-03 15:30–17:30, Promenade Level, Milano IV · Mangesh Budkule, Khurram Khawaja
Caesars Forum (3)
  • IND322-R · Chalk talk — Physical AI in Telecom: A Framework for Workload Placement

    2026-12-01 15:00–16:00, Level 1, Academy 414 · Young Jung, Angelo Sampietro, Umer Chaudhary
  • IND3304-R1 · Chalk talk — KNFSD File Cache: Extending Your NFS Storage into the Cloud

    2026-12-01 16:30–17:30, Level 1, Forum 114 · Andy Hayes, DJ Rahming
  • INV537 · Chalk talk — No torque sensors, no problem: learning safe physical interaction for humanoids

    2026-12-02 13:30–14:30, Level 1, Alliance 305 · Josh Merel
Wynn/Encore (3)
  • IND359-R1 · Chalk talk — In-Country HPC: How TGS Processes Seismic Data on AWS Brazil

    2026-12-01 12:00–13:00, Level 1, Chopin 4 · Kun Jiao, Faisal Asif
  • SEC401 · Workshop — Winning the DDoS Game: defend your app under real-time attack

    2026-12-03 12:00–14:00, Upper Convention Promenade, Cristal 5 · Jonathan Woods, Julian Garcia
  • BIZ321-R1 · Workshop — Never Start Over: Chat, Voice, Video, Screen Share on Connect Customer

    2026-12-04 08:00–10:00, Upper Convention Promenade, Cristal 1 · Ying Qian, Amit Gupta

Community Picks: sessions led by practitioners

25 practitioner-led community sessions, listed by day. These are not on Bryan's held schedule.

Monday, November 30

12:00–13:00 COM328Code talkLevel 300 – Advanced

Autonomous Remediation with OpenTelemetry and DevOps Agent

Julia Furst MorgadoJones Zachariah Noel

You instrumented everything. You have metrics, logs, and traces.

Read the rest

And when production breaks a human still has to piece it together, tab by tab, signal by signal, under pressure. That is not an observability problem, it’s an automation problem. We'll show you an architecture on AWS that closes it. OpenTelemetry captures signals across EKS, ECS, and Lambda. AWS DevOps Agent runs the investigation, finds the root cause, and prepares the fix. By the time a human looks at it, the pull request is ready. Come learn the triggers, investigation loop, and guardrails you need before letting an agent touch production.

Caesars Forum, Level 1, Forum 120, Content Hub, Code Talk

Also runs Wed Dec 2, 4:30 PM · MGM Grand, Level 3, Chairman's 363, Content Hub, Code Talk

13:30–14:30 COM302Chalk talkLevel 300 – Advanced

The judgment gap: when Terraform scans clean but fails the audit

Anton Babenko

Checkov, trivy, and Prowler check rules someone already wrote. SOC 2 CC6.1 is a sentence open to interpretation.

Read the rest

Between them sits the judgment gap, where AI-generated Terraform for S3, CloudFront, API Gateway, and Lambda often lands. We map the gap on the whiteboard together: take real controls from SOC 2 and PCI DSS, set them against what scanner rules verify, and mark what is left over. Then we discuss what closes the remainder: better authoring-side skills, human review, multi-model review, or accepting the risk. Bring a control your team owns; we trace where its enforcement lives, or does not.

Caesars Forum, Level 1, Academy 414

Also runs Thu Dec 3, 2:30 PM · Caesars Forum, Level 1, Forum 103

14:30–15:30 COM316Chalk talkLevel 300 – Advanced

Think Beyond DevOps: SRE, SOC, and Compliance in One AWS DevOps Agent

Richard FanShaoyi Li

As AI speeds up releases, continuous security control assurance and detecting threats in millions of telemetry signals get harder. We transformed AWS DevOps Agent into one agent for SRE, SOC and compliance, effectively addressing AI-era cloud security challenges in production.

Read the rest

With AWS Config Conformance Packs for compliance-as-code and security MCP for finding triage, it produces point-in-time compliance and well-architected assessments, and drives alarm-triggered agentic SOC investigations within a controlled agent space, automating responses and keeping human-in-the-loop on high-risk mitigations only. You'll learn the methodology to design an agentic compliance and SOC solution that ports seamlessly into your own environment.

Wynn/Encore, Convention Promenade, Montrachet 1

Also runs Thu Dec 3, 2:30 PM · Caesars Palace, Promenade Level, Trevi

15:00–16:00 COM326Code talkLevel 300 – Advanced

Observable Agents with Lambda Durable Functions and OpenTelemetry

Gunnar GroschRaphael Manke

When AI agents run as long-lived, multi-invocation workflows on Lambda Durable Functions, standard observability breaks down. A single durable execution spanning five specialists producing independent telemetry, replays pollute logs with duplicates, and no tool can tell you which agents are failing across thousands of concurrent runs.

Read the rest

We'll build a multi-agent purchasing coordinator with Lambda durable functions, instrument it with the OpenTelemetry plugin for one coherent trace across all invocations, and CloudWatch insights to answer fleet-wide queries with business context. Attendees leave with fully observable architecture and the investigation pattern from CloudWatch alarm to end-to-end traces and correlated log entries.

Caesars Forum, Level 1, Forum 120, Content Hub, Code Talk

15:30–15:50 COM202Lightning talkLevel 200 – Intermediate

Rebuilding a Platform's Foundation on DynamoDB Streams and Lambda

Awedis Keofteian

Our platform had single points of failure baked into its original design. In this session, I walk through how we closed those gaps across all five AWS Well-Architected pillars: moving event-driven workloads onto DynamoDB Streams and Lambda, cutting costs through smarter data tiering, and building the observability needed to trust the new design under real operational pressure.

Wynn/Encore, Upper Convention Promenade, Cristal 2, Content Hub, Lightning Theater

16:30–17:30 COM403Chalk talkLevel 400 – Expert

One-Way Doors: Launching a Regulated Financial Exchange on AWS in 2026

Margaret Valtierra

AWS turns 20 this year. Starting fresh in 2026 means navigating hundreds of services with no clear map for which decisions are one-way doors.

Read the rest

I'll share how we built a CFTC-regulated exchange and clearinghouse from the AWS Organizations up. The full security stack (ControlTower, Config, GuardDuty, Security Hub, and IAM) built in the right order from day one. We almost walked through the wrong door. A pre-launch architecture review with our engineers and AWS solution architects caught it in time. Here's how to have that conversation before it's too late.

Wynn/Encore, Convention Promenade, La Tache 2

17:30–17:50 COM343Lightning talkLevel 300 – Advanced

Craft, Don't Vibe, Your MVP

Matt Lewis

Many people today use LLMs only for code generation. The best teams embrace AI across all stages of the SDLC.

Read the rest

The decisions made before building the first version of your product directly affect its viability and success. Rather than relying on vibe coding to make these decisions implicitly, agents can be used to evaluate architectural options, compare trade-offs and challenge assumptions before a single line of code is created. This session explores how high-performing teams are investing time capturing standards, patterns and expertise, making them available for agents to apply consistently. Learn from practical real world examples how AI is being used to accelerate the delivery of production-ready systems, rather than those that carry technical debt from day one.

Venetian, Level 2, Hall B, Expo, Builder Theater

18:00–18:20 COM344Lightning talkLevel 300 – Advanced

Now serverless has servers. What now?

Evandro Pires da Silva

I spent years advocating for serverless, but as a CTO, I could not always see a viable path to bring it into certain business scenarios. In this talk, I share what has changed with the new AWS capabilities, why this can change the game for larger companies, and, along the way, I explore the concepts and how these services work in more depth to help us think about architecture with more clarity.

Venetian, Level 2, Hall B, Expo, Builder Theater

18:30–18:50 COM342Lightning talkLevel 300 – Advanced

Renaissance Dad: building an agentic learning app for my 9-year-old with Kiro

Sylvain Bruas

Werner Vogels told us to become Renaissance Developers — curious polymaths who blend disciplines and own what they build. I took it literally: as a dad, I used Kiro and agentic AI to build a learning app that helps my 9-year-old practice math and school lessons.

Read the rest

In this session, I'll share the full architecture: an OCR agent that imports exercises from photos of school notebooks, an adaptive agent that personalizes difficulty, and a coaching agent that guides instead of correcting. We'll deep-dive into multi-agent orchestration patterns, vision model integration, spec-driven development with Kiro, and the guardrails required when your end user is a child. Attendees will leave with reusable agentic patterns applicable far beyond education.

Venetian, Level 2, Hall B, Expo, Builder Theater

Tuesday, December 1

16:30–16:50 COM345Lightning talkLevel 300 – Advanced

Designing for Operability: Catching Performance Gaps Before Scale

Jofia Jose Prakash

A training workload had plenty of GPU capacity, yet utilization remained low and training time kept increasing. The GPUs were healthy.

Read the rest

The real bottleneck was upstream: the data pipeline could not feed them fast enough, leaving expensive accelerators waiting for work. This talk walks through how we traced that utilization gap, found where throughput was being lost, and identified the signals that could have exposed the problem earlier. Through this production experience, we’ll look at what it means to design for operability before scale, when performance, utilization, and cost problems are still easier to fix.

Wynn/Encore, Upper Convention Promenade, Cristal 2, Content Hub, Lightning Theater

16:30–17:30 COM329Code talkLevel 300 – Advanced

Sandboxing AI Agents on EKS: Isolation That Scales

Adit ModiIshan Modi

AI agents execute untrusted code — one prompt injection away from data exfiltration. This session covers building graduated isolation for autonomous agents on EKS using the agent-sandbox controller.

Read the rest

You'll learn why runtime tiers (runc → gVisor → Firecracker) map to different threat models, how SandboxClaim resources manage ephemeral agent lifecycle, and Cilium FQDN egress filtering that prevents exfiltration without blocking legitimate tool use. We'll whiteboard Karpenter NodePool design for gVisor-capable nodes, discuss cost tradeoffs between isolation levels, and share what breaks when you over-isolate. Leave with a production-ready isolation framework.

Caesars Forum, Level 1, Forum 120, Content Hub, Code Talk

Wednesday, December 2

10:00–11:00 COM402Breakout sessionLevel 400 – Expert

Practical PKI for On-Prem Devices with IAM Roles Anywhere

Ben Bridts

Secure authentication of your humans and machines is a challenge every organization faces, especially if your proof of identity is not already in AWS. In this session, we'll talk you through multiple ways to set up or use PKI - such as Belgium's eID or a hardware key - to securely access AWS with phishing resistant authentication built using IAM Roles Anywhere.

Read the rest

This will include live demos and code samples. By the end of the session we will stop worrying about plain text credential storage because we can use trusted hardware instead.

Wynn/Encore, Level 1, Encore Ballroom 5

12:30–12:50 COM340Lightning talkLevel 300 – Advanced

Lambda MicroVMs: The Missing Serverless Primitive for Multi-Tenant Agents

Darshit Pandya

Lambda Functions are stateless and time out after 15 minutes. Fargate tasks are expensive to keep warm.

Read the rest

When building multi-tenant agent platforms, I needed per-tenant isolation without per-tenant cost. Lambda MicroVMs solve this: Firecracker VM-level isolation, 8-hour stateful execution, auto-suspend when idle, instant resume on next request. This session walks through the architecture I built, replacing always-on containers with MicroVMs as per-tenant agent sandboxes. I'll share the tradeoffs, what surprised me, and a deployable pattern you can take home and ship.

Wynn/Encore, Upper Convention Promenade, Cristal 2, Content Hub, Lightning Theater

13:00–14:00 COM303Breakout sessionLevel 300 – Advanced

Lambda for every scale and workload shape

Liz Fong-Jones

Honeycomb spends millions of dollars a year on Lambda. Every cold query runs through a map-reduce backend that used to mean one tier, Lambda Functions, regardless of shape.

Read the rest

We outgrew Lambda Functions, not Lambda: Managed Instances now absorbs steady-state baseline traffic at roughly 7.5x lower cost than Functions, MicroVMs takes spikes of load tolerating 1.5 seconds of added latency, and Functions still catches bursty spillover. We'll walk through the observability and benchmarks that drove each cutover, and the routing logic deciding which tier a query lands on before it executes. You'll learn how to scale your serverless workloads multiple orders of magnitude without linear cost.

Caesars Forum, Level 1, Forum 120, Content Hub, Red Theater

13:30–14:30 COM404Chalk talkLevel 400 – Expert

Scaling Multi-Cluster EKS Without Burning Your Infrastructure Budget

Sagar Utekar

Managing hundreds of Amazon EKS clusters breaks standard multi-tenant architectures under the weight of shared control planes, noisy neighbors, and runaway cross-AZ data transfer costs. This interactive chalk talk skips the slides to whiteboard the painful realities of multi-cluster fleet topology at scale.

Read the rest

We will dissect the architectural trade-offs between monolithic clusters and localized micro-clusters, mapping out exact blast-radius constraints. Bring your worst infrastructure bottlenecks; we will openly debate cluster life-cycle automation, shared VPC patterns, and cost-allocation frameworks, leaving you with a battle-tested blueprint to optimize fleet operations on Monday morning.

MGM Grand, Level 1, Boulevard 155

Also runs Thu Dec 3, 12:00 PM · MGM Grand, Level 3, Premier 309

13:30–14:30 COM323Code talkLevel 300 – Advanced

From Fragmented Infra to Golden Path: An Internal Developer Platform on EKS

Johannes Koch

Every team wanted to ship faster, but each was reinventing the same wheels: clusters, pipelines, infrastructure, access. So we built an internal developer platform on Amazon EKS — one golden path that lets product teams go from idea to running service without becoming Kubernetes experts.

Read the rest

This is the honest story of that journey. We'll show how we used EKS as the foundation, Crossplane to provision cloud infrastructure through Kubernetes APIs, and Argo CD to drive everything through GitOps. We'll walk through the challenges that nearly derailed us and share what actually moved the needle. You'll leave with a realistic map of the terrain: what worked, what hurt, and what we'd do differently.

Caesars Forum, Level 1, Forum 120, Content Hub, Code Talk

16:00–17:00 COM304Breakout sessionLevel 300 – Advanced

Hidden trade-offs in distributed architectures

Anahit Pogosova

Asynchronous event-driven architectures enable loosely coupled, scalable, and resilient systems. But the devil is in the details, and hidden complexity emerges when components interact under real-world pressure.

Read the rest

Consistency, latency, failure handling, and cost aren't determined by architecture diagrams but by the trade-offs behind the integration patterns and services you choose. Through a real-world case study using AWS Lambda, Amazon DynamoDB, DynamoDB Streams, and AWS Step Functions, you'll investigate unexpected system behavior and uncover the trade-offs behind it. You'll leave with a sharper intuition to see beyond cloud abstractions and make more intentional architectural decisions for resilient systems.

MGM Grand, Level 1, Grand 117

16:30–17:30 COM322Chalk talkLevel 300 – Advanced

Eval-Driven LLM Routing: Cut Costs Without Losing Accuracy

Luca BianchiOlivier Leplus

This session shows how to build an eval-driven LLM routing system: harvest a production eval dataset, define accuracy KPIs using LLM-as-judge, set a non-negotiable accuracy floor as the practitioner's decision, benchmark candidate models, and generate a routing policy optimized for cost within that floor. Result: approximately 80% cost reduction at maintained accuracy.

Read the rest

A reference architecture you can start implementing immediately.

Caesars Forum, Level 1, Alliance 305

16:30–16:50 COM331Lightning talkLevel 300 – Advanced

AI Writes the Code. Who Reviews the Reviewer?

Rossana Suarez

AI coding assistants now write, refactor, and even review code at scale but who's actually accountable when something breaks in production? This session explores how AI has reshaped the code review process: faster first-pass reviews, AI-generated suggestions, and agentic tools running directly in CI/CD pipelines.

Read the rest

We'll examine where AI genuinely improves review quality, where it creates false confidence, and why human judgment remains the deciding factor. You'll leave with a practical framework for combining AI speed with human accountability in your own review process.

Wynn/Encore, Upper Convention Promenade, Cristal 2, Content Hub, Lightning Theater

16:30–17:30 COM319Breakout sessionLevel 300 – Advanced

Stop Rolling Back AI Agents: Production Patterns That Work

Nadia ReyhaniKristine ArmiyantsGunnar Grosch

The demo shipped. The agent went live.

Read the rest

Then someone pulled it back. 74% of organizations that have deployed an AI agent have rolled it back or shut it down, and the rate rises to 81% among those reporting fully mature guardrails. Getting agents into production got easier. Keeping them there without hurting customers did not. We'll walk through the failure patterns behind rollbacks: agents that fail silently, prompts treated as safety boundaries, and monitoring that overlooks non-deterministic failures. You'll leave with the patterns that prevent them: infrastructure-level guardrails, blast radius containment, and observability built for probabilistic systems.

Caesars Palace, Promenade Level, Roman I

Thursday, December 3

08:30–09:30 COM309Breakout sessionLevel 300 – Advanced

A DevOps Approach to Recover at Scale from Disaster or Ransomware

Samantha Eliot

Recovering your cloud workload completely after a regional event or ransomware attack is not a checklist; it demands a culture of resilience that turns individual craft into reliable recovery. That culture is built through DevOps.

Read the rest

Clear communication, team empowerment, and frequent experimentation and measurement are what separate teams that recover from teams that hope. This session covers the technical requirements for recovering cloud workloads at scale, and the DevOps practices that make those requirements actually work under pressure. You'll leave with practical knowledge to drive full recovery in the face of the worst disasters.

Caesars Forum, Level 1, Forum 120, Content Hub, Purple Theater

09:00–10:00 COM315Chalk talkLevel 300 – Advanced

Beyond the Docs: Operating Scoped VPC Endpoints in Real Environments

Kenneth AttardMeg Ashby

Scoped VPC endpoint policies look simple, until production traffic proves otherwise. In this interactive chalk talk, we'll walk through real-world failures we've encountered while deploying account- and org-scoped endpoint policies and trace the identity flows, auth chains, and hidden dependencies behind them.

Read the rest

Together, we'll explore a key lesson: endpoint policies evaluate the request that's actually made, not the request you think you're making. Bring your questions and experiences as we whiteboard the assumptions that lead to outages and learn how to spot them before they impact production.

Caesars Forum, Level 1, Forum 114

09:00–10:00 COM305Breakout sessionLevel 300 – Advanced

Why Enterprise RAG Fails: Document AI and Ontology Patterns on AWS

Bhuvaneswari SubramaniAyyanar JeyakrishnanVivek Raja P S

Most teams blame embeddings, vector databases, or foundation models when RAG systems underperform. In practice, failures often originate earlier in the pipeline where business meaning is lost during document ingestion.

Read the rest

This session explores production Document AI patterns on AWS, including layout-aware extraction, ontology-driven classification, semantic chunking, metadata enrichment, and retrieval optimization. Using financial services examples, we demonstrate how domain ontologies improve extraction accuracy, retrieval relevance, and explainability. Attendees will learn practical architecture patterns, implementation tradeoffs, and a framework for building scalable, trustworthy RAG systems for enterprise workloads.

Wynn/Encore, Upper Convention Promenade, Cristal 2, Content Hub, Purple Theater

11:30–12:30 COM314Breakout sessionLevel 300 – Advanced

Beyond the GPU count: Engineering inference at scale

Mai Nishitani

You have enough GPUs in your server to cover your software requirements. But is that all you need to consider?

Read the rest

In a world where everyone is fighting for GPUs, picking the wrong server type is one of the most expensive mistakes an engineering team makes at scale. This session shares lessons running GPU inference across AWS and bare metal connected privately. You'll see how EKS Hybrid Nodes provides a single scheduling plane, failure modes, and how to use AI to accelerate decisions without replacing engineering judgement. Leave with a cloud and bare metal architectural pattern and a GPU selection framework covering workload fit and server tradeoffs.

Caesars Forum, Level 1, Forum 120, Content Hub, Purple Theater

16:00–17:00 COM327Code talkLevel 300 – Advanced

Answering CX questions 10x faster with Amazon Connect Customer and Amazon Quick

Kobe Xu

Amazon Connect Customer generates a deep set of signals from every customer interaction. Sentiment, intent, conversation analytics, resolution outcomes.

Read the rest

It's all there and Connect Customer gives you built-in tools to explore it. But we wanted to go further. We wanted anyone in the organization to ask questions about customer interactions in plain language and get answers instantly. So we paired Connect Customer's analytics data lake with Amazon Quick. In this session, we walk through how we did it, including how we handled cross-account Lake Formation access, and show how our broader team now asks questions and reshapes queries on the fly. 10x faster from question to action.

MGM Grand, Level 1, Boulevard 159

Developer API

How this page gets its data: the AWS Events API operations, the scripts that call them, and every request and response. Raw requests stay in closed drawers.

How this page reads my AWS Events schedule

Five Events API capabilities drive this page. Each row names the endpoint and the script that calls it.

Python 3.10 or newer, no packages to install. Sign-in uses PKCE (Proof Key for Code Exchange), the browser flow the Events API requires. The access token lasts 60 minutes and the refresh token lasts 30 days. Registration for re:Invent 2026 is required for the catalog and any schedule call.

Install

git clone https://github.com/chasko-labs/reinvent26-planner.git

The clone holds the standard-library client in cli/reinvent26/api.py.

Source: docs/sign-in.md

Events API capabilities, endpoints, and scripts
CapabilityEndpointScriptWhat it gives an attendee
Full session catalogGET /v1/events/{eventId}/sessions, GET .../sessions/{sessionId}cli/reinvent26/api.pySearch every session offline by topic, then pull abstracts and seat counts for the shortlist.
FavoritesPOST /v1/events/{eventId}/favorites, DELETE .../favorites/{sessionId}cli/reinvent26/api.pyMark 10 sessions per call and see them in your schedule.
Personal calendar and scheduleGET /v1/events/{eventId}/schedulecli/reinvent26/schedule.pyOne read of favorites, reservations, and personal time, with double bookings flagged.
Personal timePOST /v1/events/{eventId}/personal-time, PUT and DELETE .../personal-time/{id}cli/reinvent26/personal_time.pyBlock meals, open windows, and travel on your own calendar.
Reserve and cancel seatsPOST /v1/events/{eventId}/reservations, DELETE .../reservations/{sessionId}cli/reinvent26/fill.pyBook open seats that fit your free windows and release one when a plan changes.

MCP

The same operations run over MCP (Model Context Protocol) at https://api.awsevents.com/mcp. The repo ships no MCP server of its own. kiro-cli config:

~/.kiro/settings/mcp.json

{
  "mcpServers": {
    "awsevents": {
      "url": "https://api.awsevents.com/mcp",
      "oauth": {
        "clientId": "7vmom55m1qstvq8i71ph127bfq",
        "redirectUri": "http://127.0.0.1:8976",
        "oauthScopes": [
          "openid",
          "email",
          "events/access"
        ]
      }
    }
  }
}

The first tool call opens the Builder ID sign-in. Ask for a schedule and the server returns favorites and reservations. Guides cover Meta Muse, Microsoft Copilot, and Google Gemini too.

Source: docs/integrations/kiro-cli.md, skills/awsevents-muse/SKILL.md, docs/mcp-approach.md

How new sessions and launches show up

The planner checks the catalog for sessions added since the last check. It only reads: it never reserves, favorites, or writes personal time.

How the watcher polls, detects, and reports new sessions
QuestionAnswer
What is polledThe full session list, 100 per page, until nextToken is missing. Then GetSession for each added session and GetSchedule once. The requests are in the box below.
How oftenOnce per run. The watcher runbook sets a timer for 07:00 and 19:00 local time; the code has no schedule of its own.
How a new session is detectedThe planner compares the sessionId values in the live list with the saved snapshot. An id in the live list and not in the snapshot is added; an id in the snapshot and not in the live list is removed. An empty live list counts as an outage, so the snapshot stays.
Where it lands on this pageEach run writes a dated report. The week export copies the added codes and the codes that fit an open seat into week.json. This page does not print them.
Product launchesThe watcher flags an added session when its title or abstract uses launch words such as new, launch, announced, what's new, preview, or generally available. That marks a session that may cover a launch. The planner does not track product launches yet: nothing reads announcements or keynote content, and this page does not print the flags.
Developer: ListSessions, GetSession and GetSchedule request and response

ListSessions, every page: request

GET https://api.awsevents.com/v1/events/reinvent2026/sessions?includeAbstracts=false&maxResults=100
Accept: application/json
Authorization: Bearer $ACCESS_TOKEN

Response, trimmed to the fields this page uses

{
  "nextToken": "$NEXT_TOKEN",
  "sessions": [
    {
      "abbreviation": "WOC102",
      "sessionId": "1787852035116001jeXM",
      "sessionTime": {
        "date": "2026-12-03",
        "length": "60",
        "time": "14:00"
      },
      "title": "The rules for AI are being written right now and nobody’s watching"
    }
  ]
}

GetSession, one call per added session: request

GET https://api.awsevents.com/v1/events/reinvent2026/sessions/$SESSION_ID
Accept: application/json
Authorization: Bearer $ACCESS_TOKEN

Response, trimmed to the fields this page uses

{
  "session": {
    "seatAvailability": "limited",
    "sessionId": "$SESSION_ID"
  }
}

GetSchedule, to check clashes: request

GET https://api.awsevents.com/v1/events/reinvent2026/schedule
Accept: application/json
Authorization: Bearer $ACCESS_TOKEN

Response, trimmed to the fields this page uses

{
  "schedule": {
    "favorites": [],
    "personalTime": [],
    "reserved": [
      "1780442252767001cb2W"
    ]
  }
}

The seatAvailability value shown is a placeholder for the enum: available, limited, veryLimited, unavailable, or walkUp.

On this page: export_week.py reads the added and opportunity codes from the latest report into the watcher block of week.json, and this page does not print them.

At re:Invent: You hear about a session added after the catalog opened while its seats may still be open.

Script: cli/reinvent26/newevents.py

Every request and response on this page

Each drawer is closed. Open one to see the HTTP request, the trimmed response, and the script that makes the call.

Calendar and events

Developer: Calendar feed request and response

GetSchedule: request

GET https://api.awsevents.com/v1/events/reinvent2026/schedule
Accept: application/json
Authorization: Bearer $ACCESS_TOKEN

Response, trimmed to the fields this page uses

{
  "schedule": {
    "favorites": [
      "1790358246739001ipkp",
      "1780441498719001G3Dy"
    ],
    "personalTime": [],
    "reserved": [
      "1780442252767001cb2W",
      "1780441496076001GRZY"
    ]
  }
}

On this page: export_week.py reads these ids, resolves each one against the session catalog, and writes week.json, the file this page reads for held sessions, shortlist rows, and open windows.

At re:Invent: Your reserved seats and your favorites arrive in one request, so you see what is booked before you plan anything else at re:Invent.

Script: cli/reinvent26/export_week.py

Developer: ListEvents and GetEvent request and response

ListEvents (no token): request

GET https://api.awsevents.com/v1/events
Accept: application/json

Response, trimmed to the fields this page uses

{
  "items": [
    {
      "eventId": "reinvent2026"
    }
  ]
}

GetEvent: request

GET https://api.awsevents.com/v1/events/reinvent2026
Accept: application/json

Response, trimmed to the fields this page uses

{
  "event": {
    "eventId": "reinvent2026",
    "eventName": "re:Invent",
    "timeZone": "America/Los_Angeles"
  }
}

On this page: eventId goes into every later path, and timeZone is why every time on this page reads as Las Vegas time.

At re:Invent: You know which clock the schedule runs on before you set a meeting time with someone outside Las Vegas.

Script: cli/reinvent26/api.py

Monday, November 30

Developer: ListSessions request and response

ListSessions, first page: request

GET https://api.awsevents.com/v1/events/reinvent2026/sessions?includeAbstracts=false&maxResults=100
Accept: application/json
Authorization: Bearer $ACCESS_TOKEN

Response, trimmed to the fields this page uses

{
  "nextToken": "$NEXT_TOKEN",
  "sessions": [
    {
      "abbreviation": "STG363",
      "level": "300 – Advanced",
      "room": "Level 3 | Room 306",
      "sessionId": "1780442252767001cb2W",
      "sessionTime": {
        "date": "2026-11-30",
        "length": "60",
        "time": "08:30"
      },
      "title": "Modernizing VMware workloads with Amazon FSx for NetApp ONTAP",
      "type": "Chalk talk",
      "venue": "MGM Grand"
    },
    {
      "abbreviation": "INV530-R1",
      "level": "500 – Distinguished",
      "room": "Level 3 | Room 350",
      "sessionId": "1780441496076001GRZY",
      "sessionTime": {
        "date": "2026-11-30",
        "length": "60",
        "time": "11:30"
      },
      "title": "EC2 at 20: Rebuilding the control plane while it is running",
      "type": "Chalk talk",
      "venue": "MGM Grand"
    }
  ]
}

On this page: Each reserved id from GetSchedule is matched to a row here by sessionId; abbreviation, title, venue, room, and sessionTime become the Time, What, and Where cells of the Monday table.

At re:Invent: Every session carries its room and start time, so you see what sits next to your seat before you walk over.

Script: cli/reinvent26/api.py

Developer: GetSession request and response

GetSession: request

GET https://api.awsevents.com/v1/events/reinvent2026/sessions/1780442252767001cb2W
Accept: application/json
Authorization: Bearer $ACCESS_TOKEN

Response, trimmed to the fields this page uses

{
  "session": {
    "abbreviation": "STG363",
    "seatAvailability": "unavailable",
    "sessionId": "1780442252767001cb2W"
  }
}

On this page: seatAvailability is the one field only GetSession returns; week.json stores it as seat, and STG363 reads unavailable there.

At re:Invent: You learn whether a session still has seats before you spend time walking to its room.

Script: cli/reinvent26/seats.py

Developer: CreatePersonalTime request and response

CreatePersonalTime: request

POST https://api.awsevents.com/v1/events/reinvent2026/personal-time
Accept: application/json
Authorization: Bearer $ACCESS_TOKEN
Content-Type: application/json

{
  "description": "Open window near MGM Grand",
  "endDateTime": "2026-11-30T11:30:00",
  "location": "MGM Grand",
  "startDateTime": "2026-11-30T09:30:00",
  "title": "Open window, morning"
}

Response

Empty body. The new personalTimeId is read back from GetSchedule.

On this page: The 09:30 to 11:30 open row in the Monday table is this window; the call writes the same span onto the attendee calendar.

At re:Invent: Free time shows as busy on your re:Invent schedule, so nothing else gets booked into it.

Script: cli/reinvent26/api.py

Tuesday, December 1

Developer: GetSchedule request and response

GetSchedule: request

GET https://api.awsevents.com/v1/events/reinvent2026/schedule
Accept: application/json
Authorization: Bearer $ACCESS_TOKEN

Response, trimmed to the fields this page uses

{
  "schedule": {
    "favorites": [],
    "personalTime": [],
    "reserved": [
      "1790358232199001iAvH",
      "1780442244219001cHBq",
      "1790358240915001ipNX"
    ]
  }
}

On this page: schedule.py turns each reserved id into a start and an end and flags any pair that overlaps; the three Tuesday rows (IND359-R1, IND322-R, IND3304-R1) come from these ids.

At re:Invent: A clash between two seats shows up before you arrive at a room you cannot sit in.

Script: cli/reinvent26/schedule.py

Developer: ListSessions request and response

ListSessions, next page: request

GET https://api.awsevents.com/v1/events/reinvent2026/sessions?includeAbstracts=false&maxResults=100&nextToken=$NEXT_TOKEN
Accept: application/json
Authorization: Bearer $ACCESS_TOKEN

Response, trimmed to the fields this page uses

{
  "nextToken": "$NEXT_TOKEN",
  "sessions": [
    {
      "abbreviation": "IND322-R",
      "level": "300 – Advanced",
      "room": "Level 1 | Academy 414",
      "sessionId": "1780442244219001cHBq",
      "sessionTime": {
        "date": "2026-12-01",
        "length": "60",
        "time": "15:00"
      },
      "title": "Physical AI in Telecom: A Framework for Workload Placement",
      "type": "Chalk talk",
      "venue": "Caesars Forum"
    }
  ]
}

On this page: iter_sessions repeats this call until nextToken is missing, because a short page is not the last page; the cards for IND322-R and the other Tuesday sessions are read from the full catalog it builds.

At re:Invent: You can browse the whole catalog, not the first hundred sessions, to find talks at the venue you are already in.

Script: cli/reinvent26/api.py

Developer: UpdatePersonalTime request and response

UpdatePersonalTime: request

PUT https://api.awsevents.com/v1/events/reinvent2026/personal-time/$BLOCK_ID
Accept: application/json
Authorization: Bearer $ACCESS_TOKEN
Content-Type: application/json

{
  "description": "Open window near Caesars Forum",
  "endDateTime": "2026-12-01T15:00:00",
  "location": "Caesars Forum",
  "startDateTime": "2026-12-01T13:00:00",
  "title": "Open window, lunch"
}

Response

No fields read. The call replaces every field of the block; omitting location sends null and clears it.

On this page: The Tuesday 13:00 to 15:00 open row is this block; a second sync sends a PUT to the stored personalTimeId instead of creating a duplicate.

At re:Invent: Changing a plan edits one calendar block, so you never end up with two.

Script: cli/reinvent26/personal_time.py

Wednesday, December 2

Developer: ListSessions request and response

ListSessions, first page: request

GET https://api.awsevents.com/v1/events/reinvent2026/sessions?includeAbstracts=false&maxResults=100
Accept: application/json
Authorization: Bearer $ACCESS_TOKEN

Response, trimmed to the fields this page uses

{
  "nextToken": "$NEXT_TOKEN",
  "sessions": [
    {
      "abbreviation": "CMP353-R",
      "level": "300 – Advanced",
      "room": "Level 1 | Boulevard 167",
      "sessionId": "1780441498886001Gubu",
      "sessionTime": {
        "date": "2026-12-02",
        "length": "60",
        "time": "10:00"
      },
      "title": "EC2 Spot: Capacity patterns and strategies from the team that runs it",
      "type": "Chalk talk",
      "venue": "MGM Grand"
    },
    {
      "abbreviation": "INV527",
      "level": "500 – Distinguished",
      "room": "Level 3 | Room 353",
      "sessionId": "1780441494886001G2fr",
      "sessionTime": {
        "date": "2026-12-02",
        "length": "60",
        "time": "10:30"
      },
      "title": "World models for indoor robots: what breaks and why",
      "type": "Chalk talk",
      "venue": "MGM Grand"
    }
  ]
}

The window filter runs in the script. The request carries no time parameter.

On this page: The slots search keeps rows whose sessionTime sits inside the Wednesday 10:00 to 13:30 window and clashes with no held seat; that is how the planner finds backups for the open row.

At re:Invent: You see which sessions fit a gap in your day without checking each start time by hand.

Script: cli/reinvent26/cli.py

Developer: CancelReservation request and response

CancelReservation: request

DELETE https://api.awsevents.com/v1/events/reinvent2026/reservations/$SESSION_ID
Accept: application/json
Authorization: Bearer $ACCESS_TOKEN

Response

No fields read. A 404 means the seat is already released, and the planner treats that as done.

On this page: A card stays under Held only while its id is in reserved, so cancelling removes the id from GetSchedule and the card drops out on the next export.

At re:Invent: A seat you will not use goes back to the pool for another attendee.

Script: cli/reinvent26/api.py

Developer: DeletePersonalTime request and response

DeletePersonalTime: request

DELETE https://api.awsevents.com/v1/events/reinvent2026/personal-time/$BLOCK_ID
Accept: application/json
Authorization: Bearer $ACCESS_TOKEN

Response

No fields read. A 404 means the block is already gone.

On this page: The Wednesday open window stays in the table above; the delete clears only the calendar entry the planner created, tracked by personalTimeId, and leaves hand-made entries alone.

At re:Invent: A block you no longer need leaves your schedule without touching anything else on it.

Script: cli/reinvent26/personal_time.py

Thursday, December 3

Developer: CreateReservations request and response

CreateReservations: request

POST https://api.awsevents.com/v1/events/reinvent2026/reservations
Accept: application/json
Authorization: Bearer $ACCESS_TOKEN
Content-Type: application/json

{
  "sessionIds": [
    "$SESSION_ID"
  ]
}

Response, trimmed to the fields this page uses

{
  "result": {
    "failed": [
      {
        "code": "sessionFull",
        "sessionId": "1780442234745001cUk8"
      }
    ],
    "successful": []
  }
}

On this page: fill.py sends one id per call and reads result.successful; a session joins reserved only when its id lands there, which is how a row becomes held in the Thursday table.

At re:Invent: A booking you can confirm: sessionFull tells you to move to the next option at once.

Script: cli/reinvent26/fill.py

Developer: ListSessions with abstracts request and response

ListSessions with abstracts: request

GET https://api.awsevents.com/v1/events/reinvent2026/sessions?includeAbstracts=true&maxResults=100
Accept: application/json
Authorization: Bearer $ACCESS_TOKEN

Response, trimmed to the fields this page uses

{
  "nextToken": "$NEXT_TOKEN",
  "sessions": [
    {
      "abbreviation": "STG412",
      "abstract": "(abstract text)",
      "level": "400 – Expert",
      "sessionId": "1780442286150001czBp",
      "title": "Find and fix storage bottlenecks for self-managed OLTP on EC2 with EBS"
    },
    {
      "abbreviation": "MAM416",
      "abstract": "(abstract text)",
      "level": "400 – Expert",
      "sessionId": "1780442282512001ctMv",
      "title": "Free your .NET app from Windows licensing with AI-powered portability"
    }
  ]
}

On this page: The shortlist keeps rows whose level starts with 400 and reads the abstract field; STG412, SEC401, and MAM416 below are all 400 Expert.

At re:Invent: An abstract lets you judge a two-hour workshop before you commit the time.

Script: cli/reinvent26/cli.py

Developer: GetSchedule request and response

GetSchedule, read personal time back: request

GET https://api.awsevents.com/v1/events/reinvent2026/schedule
Accept: application/json
Authorization: Bearer $ACCESS_TOKEN

Response, trimmed to the fields this page uses

{
  "schedule": {
    "personalTime": [
      {
        "endDateTime": "2026-12-03T15:30:00",
        "personalTimeId": "$BLOCK_ID",
        "startDateTime": "2026-12-03T14:00:00",
        "title": "Open window, afternoon"
      }
    ]
  }
}

On this page: Create returns no body, so the planner takes personalTimeId from here and uses it for later PUT and DELETE calls on the Thursday 14:00 to 15:30 window.

At re:Invent: Calendar blocks and sessions sit in one list, so you see free time and booked time together.

Script: cli/reinvent26/personal_time.py

Friday, December 4

Developer: GetSchedule request and response

GetSchedule: request

GET https://api.awsevents.com/v1/events/reinvent2026/schedule
Accept: application/json
Authorization: Bearer $ACCESS_TOKEN

Response, trimmed to the fields this page uses

{
  "schedule": {
    "favorites": [
      "1790358246739001ipkp"
    ],
    "personalTime": [],
    "reserved": [
      "1790358236486001iasR"
    ]
  }
}

On this page: findme.py sorts the reserved sessions by start and lists every gap of 60 minutes or more as an open window a visitor can use.

At re:Invent: A person who wants to meet you at re:Invent sees when you are between sessions.

Script: cli/reinvent26/findme.py

Developer: ListSessions and CreateFavorites request and response

ListSessions: request

GET https://api.awsevents.com/v1/events/reinvent2026/sessions?includeAbstracts=false&maxResults=100
Accept: application/json
Authorization: Bearer $ACCESS_TOKEN

Response, trimmed to the fields this page uses

{
  "nextToken": "$NEXT_TOKEN",
  "sessions": [
    {
      "abbreviation": "CMP318-R1",
      "services": [
        "Amazon Elastic Compute Cloud (Amazon EC2)"
      ],
      "sessionId": "1790358246739001ipkp",
      "title": "Build your iOS CI/CD pipeline with AI agents on EC2 Mac",
      "topics": [
        "Compute",
        "Developer Tools"
      ]
    }
  ]
}

CreateFavorites, optional last step: request

POST https://api.awsevents.com/v1/events/reinvent2026/favorites
Accept: application/json
Authorization: Bearer $ACCESS_TOKEN
Content-Type: application/json

{
  "sessionIds": [
    "1790358246739001ipkp"
  ]
}

Response, trimmed to the fields this page uses

{
  "result": {
    "failed": [],
    "successful": [
      "1790358246739001ipkp"
    ]
  }
}

The service list comes from your own AWS account, not from the Events API. The API has no keyword parameter in api.py: the script lists the catalog and matches keywords itself.

On this page: Services such as Amazon EC2 map to keywords in the script, sessions whose title, topics, or services match rank first, and the Friday shortlist row for CMP318-R1 comes from that ranking.

At re:Invent: You get sessions picked for the services you run, without reading thousands of titles.

Script: cli/reinvent26/cli.py

Developer: CreatePersonalTime request and response

CreatePersonalTime: request

POST https://api.awsevents.com/v1/events/reinvent2026/personal-time
Accept: application/json
Authorization: Bearer $ACCESS_TOKEN
Content-Type: application/json

{
  "description": "back at the Venetian after the session: last hallway window",
  "endDateTime": "2026-12-04T11:00:00",
  "startDateTime": "2026-12-04T10:00:00",
  "title": "Brunch"
}

Response

Empty body. The planner reads the new personalTimeId from GetSchedule and stores it for the next run.

On this page: The Friday 10:00 brunch block on the calendar is this entry; held sessions are skipped and times round to 5 minutes, starts down and ends up.

At re:Invent: Meals, windows, and travel sit on your own calendar next to your sessions.

Script: cli/reinvent26/personal_time.py

Seats

Developer: GetSession, CreateReservations and CancelReservation request and response

GetSession: request

GET https://api.awsevents.com/v1/events/reinvent2026/sessions/$SESSION_ID
Accept: application/json
Authorization: Bearer $ACCESS_TOKEN

Response, trimmed to the fields this page uses

{
  "session": {
    "abbreviation": "IND322-R",
    "seatAvailability": "veryLimited",
    "sessionId": "1780442244219001cHBq"
  }
}

CreateReservations: request

POST https://api.awsevents.com/v1/events/reinvent2026/reservations
Accept: application/json
Authorization: Bearer $ACCESS_TOKEN
Content-Type: application/json

{
  "sessionIds": [
    "$SESSION_ID"
  ]
}

Response, trimmed to the fields this page uses

{
  "result": {
    "failed": [
      {
        "code": "sessionFull",
        "sessionId": "1780442234745001cUk8"
      }
    ],
    "successful": []
  }
}

CancelReservation: request

DELETE https://api.awsevents.com/v1/events/reinvent2026/reservations/$SESSION_ID
Accept: application/json
Authorization: Bearer $ACCESS_TOKEN

Response

No fields read. A 404 means the seat is already released.

HTTP 200 on a reservation does not mean the session was booked. Read result.failed and result.successful, then confirm with GetSchedule.

On this page: The first list below holds sessions whose seatAvailability reads available, limited, or veryLimited, and the second holds sessions that are not reservable (walk-up); GetSession is the call that produces that value.

At re:Invent: You see the seat state before you walk over, and you reserve the moment a seat opens.

Script: cli/reinvent26/seats.py

Community favorites

Developer: CreateFavorites and RemoveFavorite request and response

CreateFavorites: request

POST https://api.awsevents.com/v1/events/reinvent2026/favorites
Accept: application/json
Authorization: Bearer $ACCESS_TOKEN
Content-Type: application/json

{
  "sessionIds": [
    "$SESSION_ID_1",
    "$SESSION_ID_2"
  ]
}

Response, trimmed to the fields this page uses

{
  "result": {
    "failed": [],
    "successful": [
      "1790358246739001ipkp",
      "1780441498719001G3Dy"
    ]
  }
}

RemoveFavorite: request

DELETE https://api.awsevents.com/v1/events/reinvent2026/favorites/$SESSION_ID_1
Accept: application/json
Authorization: Bearer $ACCESS_TOKEN

Response

No fields read. A 404 means the favorite is already gone.

One call takes up to 10 session ids.

On this page: The community picks above are stored as favorites, so they appear in GetSchedule under favorites and never under reserved.

At re:Invent: You mark sessions to revisit without holding a seat from anyone else.

Script: cli/reinvent26/api.py

Show me the AI

Footnote for the AI-focused attendee. The schedule above is the plan.

Learning about AI at re:Invent? These community-track sessions are worth a seat. Each one sends you home with something you can use.

Tue Dec 1, 17:00–17:20 COM339Lightning talkLevel 300 – Advanced

Cocktails to Code: Building AI Bartenders with Amazon Bedrock

Jimmy DahlqvistBuilder CenterGitHubLinkedInBlogDev.toSessionize

Plain-language AI orders fail if models hallucinate items that are out of stock. I built a reliable AI bartender grounded strictly in live inventory data.

Read the rest

This session covers how an AI agent on Amazon Bedrock uses Strands Agents, AgentCore, and MCP tools to fetch real-time menus. We will dive into the serverless architecture, see what breaks during live text streaming, and evaluate the cost versus performance tradeoffs of production grounding. You will leave with a reusable pattern and the repository to build reliable, data-driven agents.

Venetian, Level 2, Hall B, Expo, Builder Theater

You leave with a reusable pattern and the repository for turning plain-language orders into agent calls grounded in live inventory data.

Wed Dec 2, 14:30–15:30 COM308Breakout sessionLevel 300 – Advanced

What Building an SRE Agent Taught Us About Agents at Scale

Arshad ZackeriyaBuilder CenterGitHubLinkedInSiteGeethika GurugeBuilder CenterGitHubLinkedInDipin ThomasBuilder CenterSite

Anyone can wire up an agent, but operating one safely and cheaply across thousands of sessions is where developer craft begins. Grounded in a production SRE agent we built on Amazon Bedrock AgentCore and the Strands Agents SDK, this session dissects engineering around agentic models: session isolation, context management, and least privilege identity.

Read the rest

What broke, and how we fixed it: an agent with too much access, context bloat that wrecked cost, a crash during a tool call that forced durable recovery. You will see where architectures leak, how to evaluate on production traces, and tradeoffs behind resilient agent infrastructure.

Caesars Forum, Level 1, Forum 120, Content Hub, Orange Theater

You hear what broke and what held up when a team built a site reliability agent and ran it at scale.

Thu Dec 3, 16:00–17:00 COM313Breakout sessionLevel 300 – Advanced

Agentic commerce in production: multi-agent orchestration on AgentCore

Ricardo CeciBuilder CenterGitHubLinkedInSiteMatias KrederBuilder CenterGitHubLinkedInSite

A single agent that both advises and sells gets fragile fast. This session shows a production multi-agent system on Amazon Bedrock AgentCore and Strands Agents, running for real Shopify merchants, where an orchestrator coordinates two specialists: a wellness advisor in a regulated domain and a commerce agent that owns decision and checkout.

Read the rest

We walk the decisions that matter: orchestration topology, intent routing and handoffs, shared versus isolated memory across agents, where the action boundary and per-agent guardrails live, and the latency and token cost that coordination multiplies. A demo follows one conversation from advice to checkout. You leave with an orchestration pattern, and the code to implement it directly in your online store.

MGM Grand, Level 1, Grand 117

You see how a multi-agent shopping flow runs in production on AgentCore, orchestration included.

11:30 AM–12:30 PM COM325Code talkLevel 300 – Advanced

The agent orchestration spectrum: Agents, Step Functions, or Lambda?

Håkon Eriksen DrangeBuilder CenterGitHubLinkedInSite

"Orchestration mode matters more for end user experience than just model choice. In this live-coding session, I build the same GenAI workload three ways: Amazon Bedrock AgentCore, AWS Step Functions, and AWS Lambda durable functions - and show what each costs in latency, dollars, and debuggability on a production content platform.

Read the rest

Leave with a reusable decision framework to match orchestration to your workload."

Wynn/Encore, Upper Convention Promenade, Cristal 2, Content Hub, Code Talk

You get a way to pick between agents, Step Functions and Lambda for a given orchestration job.

Repeats Thu, Dec 3, 3:00 PM–4:00 PM · Caesars Forum, Level 1, Forum 120, Content Hub, Code Talk

Mon Nov 30, 11:30–12:30 COM401Breakout sessionLevel 400 – Expert

AI Didn't Wait for Security: Now What?

Ran IsenbergBuilder CenterGitHubLinkedInSiteStav OchakovskiBuilder CenterLinkedInSessionize

AI didn't wait for your security review. Developers already ship code with agentic AI, coding assistants, and MCP servers without governance; blocking them lowers efficiency and leaves you falling behind.

Read the rest

This builder-focused session dissects the architecture of a governed AI platform. You'll learn how to build a centralized MCP broker with Amazon Bedrock AgentCore Gateway and Runtime, and how agent requests gain identity and permissions. You'll build and distribute a curated agentic skills catalog and reusable agentic blueprints. Finally, you'll see how AI SDLC frameworks bring consistency and higher-quality artifacts, with DevOps and security agents woven into key stages.

Wynn/Encore, Level 1, Encore Ballroom 5

You learn which security gaps open when AI ships first, and what to fix once it is already running.